Research On Fault Prediction And Operational Safety Prediction Method For Large Wind Turbine Transmission System

In recent years,along with strengthening of people's environmental protection awareness,people pay more and more attention to the wind energy which is a renewable clean energy.Wind power generation is widely applied in the world.Fault diagnosis and operational safety prediction of the wind turbine' s components are two important factors which affect the healthy and sustainable development of the wind turbine.They have gradually become research hotspots.The transmission system is an important component in the wind turbine.The stability of its operation plays an important role in the performance of the whole wind turbine equipment.The transmission system is an important part which affects the stable operation of the wind turbine.The optimization and improvement of the transmission system fault diagnosis is of great research value to improve the efficiency of wind turbines and reduce the cost of wind turbines' operation and maintenance.It has far-reaching social significance and environmental significance for the further popularization and application of wind power generation.The characteristic frequencies of different parts in transmission system of the wind turbine are different and the difference is relatively large.In order to accurately extract various fault characteristics,it is necessary to do the classified research according to the different working conditions.At present,there is still a lack of basic theoretical research on the prediction of the operational safety of the wind turbine,and it is impossible to predict the bearing's operating state.It makes the operation and maintenance cost of the wind turbine remains high.The development of the wind power industry has been restricted.This thesis aims at the problems of the blindness of system parameter selection and the limitations of the application frequency range which are common in the application of traditional stochastic resonanc method at first.Quantum particle swarm optimization algorithm is cited.The adaptive stochastic resonance that can adjust the system parameters according to different input signals is obtained.Then the frequency information exchange method was introduced.The high frequency characteristic signal can be transferred to the low frequency reference signal where the stochastic resonance is sensitive.It solves the problem that stochastic resonance has poor effect on high-frequency signal.In addition,stochastic resonance and variational mode decomposition are combined.It enables the two methods to play their respective advantages in signal processing.The applicability and accuracy of the combination of adaptive stochastic resonance and variational mode decomposition for feature extraction are verified by simulation analysis.It provides the theoretical basis for further fault diagnosis of the wind turbine transmission system.Based on the methods of stochastic resonance,quantum particle swarm optimization algorithm and variational mode decomposition,the fault diagnosis methods for wind turbine' s motor bearing and main bearing are studied.In view of the high fault frequency of wind turbine' s motor bearing,stochastic resonance of frequency information exchange with quantum particle swarm optimization matching bearing fault diagnosis method is proposed to realize fault prognosis.The influence of wind speed variation on fault feature extraction is considered.The application range of the frequency information exchange method is be expanded.In view of the complicated background noise of the wind turbine' s main bearing,wind turbine bearing fault extraction method based on adaptive stochastic resonance de-noising and variational mode decomposition is used to realize fault prognosis.The complex optimization calculation problem which is caused by wind speed diversification is considered.Variable-scale stochastic resonance method is proposed.The efficiency of the fault diagnosis is improved.Based on time-like least squares support vector machine,unification of least squares support vector machine is proposed for the operating state prediction of the wind turbine transmission system.It not only fully considers the variation of state characteristic parameters with time,but also considers the mutual influence and constraints between different parameters.By calculating the collected sample data,it shows that the unification of least squares support vector machine method is better than the time-like least squares support vector machine and the space-like least squares support vector machine.The prediction accuracy is improved.In the process of the wind turbine transmission system prediction,the calculation result is huge and the data is inconvenient to observe.The laplacian eigenmap method and principal component analysis method are compared.The calculation results show that the laplacian eigenmap method can better retain useful information in the dimension reduction process.Finally,the laplacian eigenmap method is selected.The feature parameter matrix is reduced in dimension by this method.The fault feature can be presented in intuitive 3D graphics.The operation safety prediction of the wind turbine transmission system has been achieved.